Art Direct — AI skill for Claude Code

Art Direct

A Claude Code skill for art directing visuals. Point it at any content — an essay, a deck, a document, a webpage — and get back a creative direction you can execute.

What It Does

1. Ingests content — reads any format (HTML, PDF, DOCX, PPTX, markdown, URLs)
2. Analyzes — extracts structure, themes, narrative arc, audience, tone
3. Proposes 2-3 visual directions — each with mood, photography style, reference touchstones, and anti-cliche guidance
4. Generates section-by-section visual briefs using a Five-Lens Framework (Literal, Human, Environmental, Metaphorical, Oblique)
5. Outputs AI generation prompts — formatted for Midjourney, DALL-E, Gemini, Ideogram, and Flux
6. Generates images directly via fal.ai (Flux 2 Pro, ReCraft v3, or any supported model)
7. Critiques existing visuals — reviews images against content intent and recommends replacements

Installation

Copy the skill files to your Claude Code skills directory:

# Clone the repo
git clone https://github.com/nraford7/art-direct.git

# Copy to Claude Code skills directory
cp -r art-direct ~/.claude/skills/art-direct

The skill will be available as `/art-direct` in Claude Code.

Usage

/art-direct                          # Full workflow — ingest, propose, brief
/art-direct --style editorial-warmth # Apply a house style template
/art-direct --critique               # Review existing visuals against content
/art-direct --section "concept"      # Quick single-section visual brief

Image Generation

When you select "Generate now," the skill generates images via the fal.ai API. Requires `$FAL_API_KEY` environment variable.

Supported models:

• Flux 2 Pro (fal-ai/flux-pro/v1.1) — photographic realism
• ReCraft v3 (fal-ai/recraft-v3) — high-res photorealistic
• Ideogram v2 (fal-ai/ideogram/v2) — text-in-image, conceptual

Typical cost: ~$0.05-0.10 per image.

House Style Templates

Reusable visual directions stored as YA